diff --git a/fabtests/common/hmem_cuda.c b/fabtests/common/hmem_cuda.c
index 2f02b6f474c..e4aef962fb6 100644
--- a/fabtests/common/hmem_cuda.c
+++ b/fabtests/common/hmem_cuda.c
@@ -157,9 +157,9 @@ int ft_cuda_init(void)
 		goto err;
 	}
 
-	cuda_handle = dlopen("libcuda.so", RTLD_NOW);
+	cuda_handle = dlopen("libcuda.so.1", RTLD_NOW);
 	if (!cuda_handle) {
-		FT_ERR("Failed to dlopen libcuda.so\n");
+		FT_ERR("Failed to dlopen libcuda.so.1\n");
 		goto err_dlclose_cudart;
 	}
 
diff --git a/src/hmem_cuda.c b/src/hmem_cuda.c
index 0580bdb24f1..ec626bdada5 100644
--- a/src/hmem_cuda.c
+++ b/src/hmem_cuda.c
@@ -487,22 +487,17 @@ static int cuda_hmem_dl_init(void)
 		return -FI_ENOSYS;
 	}
 
-	cuda_attr.driver_handle = dlopen("libcuda.so", RTLD_NOW);
+	cuda_attr.driver_handle = dlopen("libcuda.so.1", RTLD_NOW);
 	if (!cuda_attr.driver_handle) {
 		FI_WARN(&core_prov, FI_LOG_CORE,
-			"Failed to dlopen libcuda.so\n");
+			"Failed to dlopen libcuda.so.1\n");
 		goto err_dlclose_cuda_runtime;
 	}
 
-	cuda_attr.nvml_handle = dlopen("libnvidia-ml.so", RTLD_NOW);
+	cuda_attr.nvml_handle = dlopen("libnvidia-ml.so.1", RTLD_NOW);
 	if (!cuda_attr.nvml_handle) {
-		FI_INFO(&core_prov, FI_LOG_CORE,
-			"Failed to dlopen libnvidia-ml.so.  Trying libnvidia-ml.so.1\n");
-		cuda_attr.nvml_handle = dlopen("libnvidia-ml.so.1", RTLD_NOW);
-		if (!cuda_attr.nvml_handle) {
-			FI_WARN(&core_prov, FI_LOG_CORE,
-			"Failed to dlopen libnvidia-ml.so or libnvidia-ml.so.1, bypassing nvml calls\n");
-		}
+		FI_WARN(&core_prov, FI_LOG_CORE,
+			"Failed to dlopen libnvidia-ml.so.1, bypassing nvml calls\n");
 	}
 
 	CUDA_DRIVER_FUNCS_DEF(CUDA_DRIVER_FUNCS_DLOPEN)